Clustering Method for Fast Contentbased Image Retrieval

Abstract

Abstract:When very large image families are involved in query processes, methods of content-based image retrieval must be optimized with a goal function determining a computing complexity. A clustering method which at the image retrieval stage ensure minimal number of comparisons of a query image and images from image database is proposed. Clustering can be fulfilled in feature or signal space. Pointwise set maps are used as the tools to find required partitions.Key words: image retrieval; clustering; number of matches; pointwise set map. INTRODUCTIONSoftware and hardware development of image acquisition, storage, proc essing, recognition, display and communication produces a background to make image databases affordable and what is more prevalent conformably to a number of applications. Highest information capacity of imagery data emphasizes barest necessity of advanced research as well. However the effecttiveness of the content-based image retrieval (CBIR) depends on chosen feature space, as evaluation of image similarities in database and a query image is computationally expensive. Moreover, when very large image collections are in use, even though reduced dimensions of high-level features do not ensure efficient query processing. One of widespread approaches to minimize time outlay consists inapplicationofvarioustechniquesof preliminary processing of image database including clustering methods 1,2,3,4,5,6 and so mediated or direct matching is one of the central problems in CBIR. Such 946 K. Wojciechowski et al. (eds.), Computer Vision and Graphics,[946][947][948][949][950][951][952]

Multi-Level Search Space Reduction Framework for Face Image Database

“…However, the computational time of vertical reduction increases linearly with dimension of data. Kinoshenko et al (2006) developed ɛ clustering technique for fast content based image retrieval. The major motive behind the work is to minimize the large number of templates involved in the comparison process thereby guaranteeing minimal number of matches.…”

Contribution-based clustering algorithm for content-based image retrieval

“…Organizing the retrieved search results into clusters is an intuitive form of content representation [14] and facilitates user's browsing of images [15]. Image clustering can also be used to optimize the performance of a CBIR system [16]. While the performance of a number of clustering algorithms in image retrieval have been analyzed in previous works [17,18,19,20], we apply our proposed algorithm to CBIR and compare its performance with that of the k-means clustering algorithm.…”